Using Ocean Data to Predict Monthly Air Temperatures of California's Coastal Cities

Using Ocean Data to Predict Monthly Air Temperatures of California's Coastal Cities | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Using Ocean Data to Predict Monthly Air Temperatures of California's Coastal Cities Xiaolin Zhang, Ming Feng This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4577041/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Air temperature anomalies in the narrow California coastal boundary layer is closely associated with the adjacent coastal sea level variability. Coastal sea levels, proxies of upper ocean heat content, are regulated by alongshore winds and poleward propagating coastal Kelvin waves. Increased upper ocean heat content or higher sea level can enhance oceanic heat release to the atmosphere which will be carried inland due to the prevailing onshore winds. Based on these physical processes, sea level anomaly (SLA) and sea surface temperature (SSTA) are combined and used as a predictor here to forecast air temperature anomaly at 4 main cities along the California coast. Results show that air temperature in these land-based stations can be predicted by SLA and SSTA with 2–3 months in advance. The prediction capability is better during late summer/early fall when the upwelling is strongest. The predicting ability is limited within a 2–3 degree narrow coastal region, and it decreases rapidly for inland stations. Our work also suggests that it is very important to enhance the resolution of models near coastal region to improve seasonal prediction. Ocean Dynamics Air Temperature Prediction California's Coastal Cities Sea Level Anomaly Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Key Points (1) A new predictor based on sea level and sea surface temperature anomaly is postulated for predicting air temperature at 4 main cities along California coast. (2) Sea level and sea surface temperature can predict air temperature with 2-3 months in advance. The prediction capability is better during late summer/early fall. (3) The predicting ability is limited within a 2-3 degree narrow coastal region and it decreases rapidly for inland stations. Plain Language Summary This study postulates using Sea Level and Sea Surface Temperature Anomaly (SLA and SSTA) to predict the air temperature anomaly at 4 main coastal cities in California. Our results suggest that air temperature can be predicted by SLA/SSTA with 2-3 months in advance. The prediction capability is better during late summer/early fall, when the upwelling is strongest. The predicting ability is limited to a 2-3 degree coastal region, and it decreases rapidly further inland. Our work also suggests that it is very important to enhance the resolution of models near coastal region to improve seasonal prediction. 1. Introduction Advances in summertime temperature forecasts are important for planning in different economic sectors, especially the energy industry. This issue is especially important in California where the summer heavy loads from air conditioners, pumping water and other seasonal issues drive the peak energy demand 50% higher than that in winter (Alfaro, EJ et. al, 2004). Barnett and Preisendorfer (1987) explored the predictability of surface air temperature over the United States with advanced statistical techniques designed to maximize predictive skill and they checked the ability of a form of persistence, Sea Surface Temperature (SST) and Sea Level Pressure (SLP) fields. They also mentioned the potential role of El Niño-Southern Oscillation (ENSO), which was unclear at that time. Alfaro EJ et al. (2004) discussed a statistical prediction model based on canonical correlation analysis (CCA) and used the spring Pacific Sea Surface Temperature (PSST) as a predictor. They also mentioned that the Pacific Decadal Oscillation index (PDO) might be used as a potential predictor. But the correlation of their prediction model is not very high, the skill of seasonal average temperature (T mean ) and cooling degree days (CDD) prediction shows values greater than 0.4 (more than 20% of the variability predicted by model) in a large part of the region, while the prediction of warm extreme values provides acceptable values only for some locations in California (see Fig. 2 in Alfaro EJ et al., 2004) and only about 30% of the stations, the correlation is greater than 0.4 (see Fig. 2 b in Alfaro EJ et al., 2004) ; furthermore, their study was mainly focused on the warm temperature anomalies and their spatial distribution, without detailed information related to the main cities along the California coast. Dottori and Clarke (2009) show that low-frequency upper-ocean temperature anomalies near the California coast are mainly due to the advection of the mean temperature gradient by anomalous alongshore/vertical flow associated with coastal Kelvin waves and westward propagating large-scale Rossby waves. These waves are associated with coastal upwelling and sea level anomalies. Since both the anomalous vertical and alongshore currents are proportional to the time derivative of the interannual sea level (Dottori and Clarke, 2009), the upper ocean heat content anomaly (and SST) are closely linked to and in phase with the sea level anomaly. Thus, both of SST and sea level can be used to predict air temperature anomaly in the California coastal cities, as the predominant onshore winds along the coast can bring anomalous heat flux at the air-sea interface toward the continent (Morales-Acuña et al., 2019).Sea level variations may represent the integrated effects of both remote and regional climate variability. Thus, a combination of SST and sea level may better capture the oceanic influences on land temperatures and serve as predictors. Land surface temperature predictions have been attempted by seasonal forecast models (Tan et al., 2014, Yang et al., 2022). We plan to assess the prediction skills with a simple statistical model in comparison with a fully coupled ocean-atmosphere-land model. Similar approaches have been adopted in other regions (Clarke and V. Gorder, 2001, 2003). The rest of the study is organized as follows. The data and methods used are summarized in Section 2 . In Section 3 we show that sea level and sea surface temperature data can be used to predict air temperature 2–3 months in advance, and it is better during late summer/early fall in 4 main cities, when the upwelling is strongest. However, the prediction is only limited within a 2–3 degree coastal region; for stations further inland, the prediction skill declines rapidly. Moreover, a simple statistical model is set up, which can predict May to August T air better with 3.5 to 5 months in advance at San Diego Lindbergh Field. Section 4 discusses the potential link of SL (SST) prediction with ENSO dynamics and concludes this study with remarks. 2. Data and methods 2.1. Dataset In this study, we focus on four clusters of stations along the California coast (Fig. 1 , Table S1 ). The “Revised Local Reference” in situ Sea Level (SL) data contains tide gauge observations, available from Permanent Service for Mean Sea Level website (PSMSL, https://www.psmsl.org/data/obtaining/ ); the land based monthly air temperature (T air ) station data is downloaded from Global Historical Climatology Network (GHCN)-Monthly Summaries ( http://www.ncdc.noaa.gov/cdo-web/datatools/findstation ), which is a database for historical monthly temperature, precipitation, and snow records over global and areas. Monthly anomalous 1°x1° Sea Surface Temperature (SST) grid data (89.5 o S to 89.5 o N, 0.5 o E to 359.5 o E) from 1960 to current is available from the International Comprehensive Ocean-Atmosphere Data Set (ICOADS, http://www.esrl.noaa.gov/psd/data/gridded/data.coads.1deg.html ). The 0.25°x0.25° topography data is from Joint Institute for the Study of the Atmosphere and Ocean (JISAO) data set ( http://research.jisao.washington.edu/data_sets/elevation/ ) and Global Ocean Data Assimilation System (GODAS) data is from ( https://www.cpc.ncep.noaa.gov/products/GODAS/ ). For each cluster, the in-situ SL data is available at the ocean station, air temperature data (T air ) is available at each ocean station and four inland stations. The SST is based on ICOADS gridded data, and a grid closest to the in-situ ocean station is chosen. The correlation is estimated based on the overlap period of the data sets. For all the observed data, the monthly anomaly is calculated by removing the climatological mean of each calendar month throughout the record. 2.2. Cross-validated correlation skill The data is separated into training and validation periods, here we used the period of January 1982 -December 2006 for training and 2007 January-December 2013 for validation. To evaluate the predictability of SLA and SSTA, here we used cross-validated correlation skill. As defined in Clarke and Van Gorder (2003), the correlations are estimated from the “Start Month” to a later month of Δt. For example, the correlation skill for a prediction of the March value of T air from January to a month of Δt = 1.5months later, i.e., the middle of March. 3. Results 3. 1. Prediction Results In order to see the potential capability of using SST/SL to predict T air at the four clusters of land-based stations, the correlation between monthly anomalous SST vs T air and SL vs T air of the coastal stations of the four clusters are first assessed (Fig. 2 , also see the first column in Table S2). Our analysis suggests that the maximum correlation is 0.81 (R crit (95%) = 0.41) between SST and T air during September at San Diego Lindbergh Field (S 1,0 ), and 0.86 (R crit (95%) = 0.39) between SL and T air during September. At Santa Monica Pier (S 2,0 ), the maximum correlation between SST (SL) and T air also occurs during September, with a correlation of r = 0.84 (R crit (95%) = 0.42) for SST and r = 0.72 (R crit (95%) = 0.40) for SL. At Watsonville Waterworks (S 3,0 ), the maximum correlation between SST(SL) and T air is slightly lower than S 1,0 and S 2,0 , which may be due to the topography of this region (see Fig. 1 ). At the last ocean station (S 4,0 , Eureka Weather Forecast Office Woodley Island), the maximum correlation between SST and T air occurs in August (r = 0.87, R crit (95%) = 0.47), and happens in May between SL and T air (r = 0.73, R crit (95%) = 0.45). Here the critical value is estimated based on Ebisuzaki et al. (1999) and the correlation between SST (SL) and T air is estimated based on the overlapped time period of two data sets. From this figure, we can see that during late summer/early fall (May to September), the correlation between SST, SL and T air are the highest. In other words, the influence of SST (SL) on T air appears to be strongest during late summer/early fall. Physically, this is mainly due to that the upwelling-favorable winds are the strongest during late summer/early fall, with the onshore component of the wind brings ocean induced temperature anomalies inland. At most coastal stations, the correlations between SL (or SST) and T air are the highest during this time of the year. Note here that, the correlation between SL (SST) and T air is largest for S 1,0 (San Diego Lindbergh Field) and decreases at the other three stations. In order to assess the potential lead months that SST (SL) can predict T air during late summer/early fall at each ocean station, lead correlations of SST with T air are calculated at each station (see Fig. 3 ). For example, Fig. 3 a is the lead correlation between SST and T air during July-September. Our results suggest that SST can be used to predict T air during July-September for a lead of about 1 month (correlation decreases to about 0.4 after 1 month). However, the lead correlation between SL and T air suggests that SL can be used to predict T air by 2 months lead (correlation is about 0.4 after 2 months) in July-September. Physically, this is because the sea level is related to upper ocean heat content, reflecting both remote and regional climate drivers. Similarly, the lead correlations of SST (SL) with T air at the other three stations are estimated during the late summer/early fall (Fig. 3 c-h). The results suggest that at the three stations, SL can predict T air with a longer lead than SST. In other words, SL is a better predictor than SST at all four stations. More specifically, at S 2,0 , SST can predict T air by 1 month, however, SL can predict T air by 3months. At S 3,0 , SST can predict T air by 1 month, SL can predict T air by 3 months. At station S 4,0 , SST can predict T air by 1 month and SL can predict T air by 2 months. In summary, for all four cities, SST leads T air with a correlation of 0.5 or greater for some lead of about 1 month (Fig. 3 , (a), (c), (e)). In contrast, for SL, the lead increases to 2–3 months (Fig. 3 , (b), (d), (f)). To explore the coastal region (coastal boundary layer) where the sea level can provide prediction of T air , near each alongshore ocean station, we picked up more stations inland as listed in Table S1 and calculated the maximum correlation for each calendar month. Figure 4 shows that the correlation between monthly anomalous air temperature and SST (SL) falls with distance from the coast. Since the correlation falls very rapidly, such a marine boundary layer is difficult to be resolved by numerical models with a coarse resolution. Also note that the sudden drop in correlation for Fig. 4 c and d (the dramatically lower correlation between SST and T air ) might be associated with land topography, which blocks the influence of the ocean. The substantial correlation in Fig. 4 d for Eureka far inland may have a topographic reason (see Fig. 1 ). Table S2 further shows the maximum correlation at each station, which imply that the maximum correlations occur in late summer and fall (August-September) at all the stations (see second column in Table S2). 3.2. Cross-validated correlations Based on the precursor properties of T air , SL and SST, we write our predictor of T air (t + Δt) from month t as the linear sum of T air , SST (SL) as follows (also see Clarke and Van Gorder, 2003) T air (t + Δt) = αT air (t)+βY(t) (1) where T air is the raw air temperature value, Y(t) is the raw sea surface temperature value (or sea level value). In order to test the ability of (1) to predict T air (t + Δt) for various leads Δt, we determined the coefficients α and β by least-square fitting over a training (validation) period for each time Δt, here, we determined the coefficients over January 1982 to December 2006 (January 2007 to December 2013). Anomalies used during a given training period were based on seasonal cycles estimated during the whole period (January 1982 to December 2013). Cross-validated correlations for various lead times Δt at San Diego Lindbergh Field are shown in Table.S3. A lead time of ½ month corresponds to (say) predicting the October value of T air given data to the end of September. Similarly, predicting the November T air value from the end of September would correspond to Δt = 1.5months, etc. Figure.S1 shows that the model predicts T air in May to August with lead of 3.5 to 5 months. 3.3. Improving the models In order to improve the prediction, we combine sea level and sea surface temperature as shown in (2) T air (t + Δt) = αT air (t)+βSST(t)+γSL(t) (2) Similar to (1), we determined coefficients α, β and γ by least-square fitting over a training period for each time Δt. Figure 5 shows correlation skill at San Diego Lindbergh Field as a function of forecast lead for “prediction” of T air using model (2) (blue solid curve), the model with only sea level (green solid curve) or SST (red solid line). The skills are cross-validated for the period of January 1982 to December 2006 for training and January 2007 to December 2013 for testing, we can see that the sea level can be used as a predictor better than sea surface temperature, and combining sea level and SST is the best. The persistence skill (black dashed line) suggests that the T air can predict itself by about 1 month with a correlation of about 0.82. Note that the lead correlation based on calendar month (Fig. 3 a, b) suggests that comparing to SST, SL can predict T air by 2 to 3 months instead of 1 month. Here we chose San Diego Lindbergh Field, because Figs. 2 and 3 suggest that at this station, the link between SL and SST is the strongest among all four ocean stations. In order to test the predictability of ENSO, we also plotted the predicting result based on ENSO index (Nino3.4), the yellow line in Fig. 5 , we can see that Nino3.4 has similar predictability to SST. Since the anomalous sea level at Santa Monica Pier is highly correlated at zero lag with the anomalous sea level at San Diego Lindbergh Field (not shown). We could also use the sea level at San Diego Lindbergh Field back to 1906 to predict the air temperature at the station near Santa Monica using Eq. (1) with no SST term. We could even try to add in the SST term using the monthly SST results at Scripps Pier from Rasmussen et al. (2019). In order to evaluate the performance of air temperature prediction near station San Diego Lindbergh, we also estimated the root mean square error (RMSE) between observed T air and the predicted results based on SST, sea level, Nino3.4 index, and the combination of sea level and SST and the persistent skill (see Fig. 6 ). The results show that the RMSE for the SST-based prediction is larger than that based on SL, and the RMSE based on combining SL and SST is the smallest, which is consistent with Fig. 5 . In Fig. 7 , we show model skills stratified into different months, based on the model with Nino3.4 index and SL as predictors (T air (t + Δt) = αT air (t)+βNino3.4(t)+γSL(t) ), we can see that during May-July, the combination of Nino3.4 and SL can predict T air with a lead about 4 months. 4. Concluding remarks and discussion Using sea level, sea surface temperature and air temperature data available at ocean stations, a statistical model based on sea level and sea surface temperature was developed to predict the air temperature in the California coast region. The analysis shows that the sea level is a better predictor than SST and it also works better for late summer\early fall than other seasons. Among four cities, the air temperature in San Diego Lindbergh can be predicted better. The prediction ability is limited to a 2–3 degree narrow “boundary layer” along California coast, which is difficult to resolve for numerical model with coarse resolution. We also noticed that the sea level can predict itself very well; for example, at San Diego Lindbergh, sea level can predict itself by 2–3 month lead with a correlation about 0.7 during July to October (see Figure S1 ). The sea level and SST can be used as air temperature predictor is because physically, \(\partial T{\prime }/\partial t\) is proportional to \(\partial \eta {\prime }/\partial t\) and integration with respect to time. The relationship that \(\partial T{\prime }/\partial t\) is proportional to \(\partial \eta {\prime }/\partial t\) follows from the balance at the sea surface (see Eq. 3.6 in Dottori and Clarke) and v' is proportional to \(\partial \eta {\prime }/\partial t\) from the freely propagating long wave Rossby wave balance. The San Diego sea level is basically in phase with the equatorial sea level, because even if the coastline were roughly 6000km from the equator to San Diego, it would only take about a month for the coastal Kelvin wave to cover this distance. Intra-seasonal Kelvin waves (periodicity about 30 days-75 days) do propagate along that coastline, and a monthly sea level would have some of that intra-seasonal signal contained in it. The 3-month running mean "seasonal" signal, which would filter out the intra-seasonal signal, may have a slightly different lag behavior. The lag due to a frictionless Kelvin wave propagation may be slight, anyway. Dissipation of the equatorial signal can also result in poleward propagation. Therefore, we expect that the persistence properties of \(\eta\) at San Diego with the ENSO \(\eta\) persistence properties at the equator. Also notice here, as discussed in the introduction, previous work has suggested the potential role of ENSO (Barnett and Preisendorfer, 1987) in predicting air temperature along California; however, the related dynamics are still unclear. For instance, ENSO generally has a peak in winter, however, our model suggests that SST and SL are highly correlated with T air during late summer/early fall. Basically, the ENSO dynamics can influence on air temperature prediction along the California coastal cities in the following two ways. Firstly, as shown in Dottori and Clarke (2009), the low-frequency upper-ocean temperature anomalies near the California coast are mainly due to the advection of the mean temperature gradient by anomalous alongshore (v’) and vertical (w’) flow associated with the westward propagating large-scale Rossby waves. These waves are associated with coastal upwelling and sea level anomalies, and they can be traced back to the coastal Kelvin waves remotely driven by the equatorial dynamics instead of local forcing (Enfield and Allen 1980; Chelton and Davis 1982; Kessler 1990). The upwelling is strongest during late summer/early fall, when the correlation between T air and SL is highest. Furthermore, the ocean dynamics basically sets the stage of T air . Along the coast of California, the climatological wind is equatorward, and it drives the cold coastal upwelling. The local wind stress can change due to variability in the atmospheric circulation. The other important factor is the ocean dynamics associated with ENSO, and these perturbations are mostly originated from the equatorial region and move poleward along the coast. The amplitude and duration of these Kelvin wave groups can enhance/reduce the coastal upwelling and change the local ocean condition in terms of SST and SL. The SST/SL anomaly for each month over the past 40 years based on GODAS data suggests that they are closely related to ENSO events, and they can be traced back to the equatorial region (not shown). Overall, our study helps to improve our understanding of the seasonal prediction along California coastal region and suggests that it is very important to enhance the resolution of models near coastal region to improve seasonal prediction. Declarations Ethics approval and consent to participate Not applicable Consent for publication Not applicable Availability of data and materials The in situ Sea Level data is available from PSMSL (https://www.psmsl.org/data/obtaining/) . The land based monthly air temperature station data is downloaded from Global Historical Climatology Network (GHCN)-Monthly Summaries (http://www.ncdc.noaa.gov/cdo-web/datatools/findstation). Monthly anomalous 1°x1° SST data is available from the International Comprehensive Ocean-Atmosphere Data Set (ICOADS) (http://www.esrl.noaa.gov/psd/data/gridded/data.coads.1deg.html). The topography data is from JISAO data set (http://research.jisao.washington.edu/data_sets/elevation/). Competing interests The authors declare that they have no competing interests. Funding This research is supported by Dalian Maritime University Start Up funding and the Fundamental Research Funds for the Central Universities (3132024123). Authors' contributions - provide individual author contribution X.Z. and M.F. wrote the main manuscript text and X.Z. prepared figures 1-7. All authors reviewed the manuscript. Acknowledgement: We gratefully acknowledge support from Dalian Maritime University Start Up funding and the Fundamental Research Funds for the Central Universities (3132024123). The original idea was based on XZ’s PhD work at Florida State University. Prof. Allan J. Clarke first suggested to predict air temperature along California coast by using sea level data. Thanks also goes to Dr. Ruixin Huang, Mikhail Karpytchev, Takashi Mochizuki, whose comments have improved the original manuscript significantly. References Alfaro E J, Gershunov A, Cayan D R (2004) A method for prediction of California summer air surface temperature. EOS Trans. AGU 85:553: 557–558. Barnett T P, Preisendorfer R (1987) Origins and levels of monthly and seasonal forecast skill for united states surface air temperatures determined by canonical correlation analysis. Mon Wea Rev. 115: 1825–1850. Chelton D, Davis R (1982) Monthly mean sea-level variability along the west coast of North America. J Phys Oceanogr. 12: 757–784. Clarke A J, Lebedev A (1999) Remotely Driven Decadal and Longer Changes in the Coastal Pacific Waters of the Americas. Journal of Physical Oceanography. 29: 828–835. Clarke A J, Dottori M (2008) Planetary Wave Propagation off California and Its Effect on Zooplankton. J ournal of Physical Oceanography. 38: 702–714. Clarke A J, Van Gorder S (2001) ENSO prediction using an ENSO trigger and a proxy for western equatorial Pacific warm pool movement. Geophysical Research Letters. 28(4): 579–582. doi:10.1029/2000GL012201. Clarke A J, Van Gorder S (2003) Improving El Niño prediction using a space-time integration of Indo-Pacific winds and equatorial Pacific upper ocean heat Content. Geophysical Research Letters. 30(7), doi:10.1029/2002GL016673. Dottori M, Clarke A J (2009) Rossby Waves and the Interannual and Interdecadal Variability of Temperature and Salinity off California. Journal of Physical Oceanography. 39: 2543–2561. Ebisuzaki W (1997) A method to estimate the statistical significance of a correlation when the data are serially correlated. J Climate. 10: 2147–2153. Enfield D B, Allen J S (1980) On the structure and dynamics of monthly mean sea level anomalies along the Pacific coast of North and South America. Journal of Physical Oceanography. 10(4): 557– 578, doi:10.1175/1520-0485(1980)0102.0.CO;2. Morales-Acuña E., Torres C R, Linero-Cueto J R (2019) Surface wind characteristics over Baja California peninsula during summer. Reg. Stud. Mar. Sci., 29, p. 100654, 10.1016/j.rsma.2019.100654. Kessler W S (1990) Observations of long Rossby waves in the northern tropical Pacific. Journal of Geophysical Research: Oceans. 95 (C4): 5183–5217. Phan-Van T, Nguyen-Quang T, Trinh-Tuan L ,et al. Seasonal Prediction of Summer near-surface Air Temperature for Vietnam using RegCM4.2[C]//The Third International MAHASRI/HyARC Workshop on Asian Monsoon and Water Cycle.2013.DOI:10.5930/issn.1994-4683.2014.07.113.p168-171. Rasmussen L L, Carter M L, Flick R E, Hilbern M, Fumo J T, Cornuelle B D, et al. (2020) A century of Southern California coastal ocean temperature measurements. Journal of Geophysical Research: Oceans. 125, e2019JC015673. https://doi.org/10.1029/2019JC015673. Yang Y, Sun W, Zou M, Qiao S and Li Q (2022) Multi-model seasonal prediction of global surface temperature based on partial regression correction method. Front. Environ. Sci. 10:1036006. doi: 10.3389/fenvs.2022.1036006. Additional Declarations No competing interests reported. Supplementary Files SupplementaryMaterials20240613.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-4577041","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":316406908,"identity":"ad953ff8-8de7-4415-ae48-c55e0b56cbda","order_by":0,"name":"Xiaolin Zhang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAz0lEQVRIiWNgGAWjYFACxsYHEhVAmvkAiMdMjBbmZgOJM0CaLYFoLextEoxtpGgxOH+wQcJy3rbEBjb2ZxIMFdaJDYS0SDYcbDCQ3HYbqIXHTILhTDphLfyMjQ0JYC3yPWxAFx4mrIWNmbHhgOSc2xCHMf4jQgs/G9AayQaQFgYzCcYGIrRI9jA2M0gcu23cxsZjbJFwLN2YoBaD88ef/5aouS3bz8b+8MaHGmtZglpAgFmCgcERFDUMCcQoBwHGDwwM9sQqHgWjYBSMghEIAOYPO5R6OjM3AAAAAElFTkSuQmCC","orcid":"","institution":"Dalian Maritime University","correspondingAuthor":true,"prefix":"","firstName":"Xiaolin","middleName":"","lastName":"Zhang","suffix":""},{"id":316406909,"identity":"1c151ba1-49e7-4084-b49c-647abd02d270","order_by":1,"name":"Ming Feng","email":"","orcid":"","institution":"Environment, Commonwealth Scientific and Industrial Research Organization (CSIRO), Indian Ocean Marine Research Centre, WA, Australia","correspondingAuthor":false,"prefix":"","firstName":"Ming","middleName":"","lastName":"Feng","suffix":""}],"badges":[],"createdAt":"2024-06-13 15:02:33","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4577041/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4577041/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":59749230,"identity":"6f81da15-462d-4e1c-8597-d5af11b0a6b0","added_by":"auto","created_at":"2024-07-05 19:17:07","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":122172,"visible":true,"origin":"","legend":"\u003cp\u003eLocations of four station clusters. Margent squares indicate the ocean stations along the California coast: OS1 (San Diego Lindbergh Field), OS2 (Santa Monica Pier), OS3 (Watsonville Waterworks), OS4 (Eureka Weather Forecast Office Woodley Island); SST, air temperature and sea level data are collected in these ocean stations. The other four land-based stations for each station cluster are depicted by red diamonds, where the air temperature data is collected. More details about each station are shown in Table S1.\u003c/p\u003e","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-4577041/v1/6e25859a860fd997e380db27.jpeg"},{"id":59749231,"identity":"12536274-a7ef-403a-8699-8a5f32e025d5","added_by":"auto","created_at":"2024-07-05 19:17:07","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":54698,"visible":true,"origin":"","legend":"\u003cp\u003eCorrelation between air temperature and SST (black line) or sea level (red line) for each month at four ocean stations along the California coast. The dashed line indicates 90% significance test based on Ebisuzaki (1997).\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-4577041/v1/023fd05457444c6c3eecbcc8.png"},{"id":59749233,"identity":"703af10c-e557-47ed-8fe6-d81940db654a","added_by":"auto","created_at":"2024-07-05 19:17:07","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":156508,"visible":true,"origin":"","legend":"\u003cp\u003eLead correlation of SST and T\u003csub\u003eair\u003c/sub\u003e (first column) , SL and T\u003csub\u003eair \u003c/sub\u003e(second column) and T\u003csub\u003eair\u003c/sub\u003e and T\u003csub\u003eair \u003c/sub\u003e(third column) for July-September at the four station clusters. July (black line), August (blue line) and September (red line).\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-4577041/v1/fc6359f1e559c08feb9bcf3e.png"},{"id":59749232,"identity":"1e2f14f3-e713-447a-a516-4bd37f012292","added_by":"auto","created_at":"2024-07-05 19:17:07","extension":"jpeg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":77881,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eMaximum \u003c/strong\u003ecorrelation between monthly anomalies of T\u003csub\u003eair\u003c/sub\u003e vs SST (black line) and T\u003csub\u003eair \u003c/sub\u003evs SL (red line) for four station clusters. The latitude and longitude of each station are listed in Table S1.\u003c/p\u003e","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-4577041/v1/1f290e88c84e1c1616286abb.jpeg"},{"id":59749236,"identity":"c4e3e7b0-f677-4a1a-b708-ab12ab0824e0","added_by":"auto","created_at":"2024-07-05 19:17:07","extension":"jpeg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":74929,"visible":true,"origin":"","legend":"\u003cp\u003eCorrelation skill as a function of forecast lead for “prediction” of T\u003csub\u003eair\u003c/sub\u003e using\u003c/p\u003e\n\u003cp\u003emodel (2) (blue solid curve, T\u003csub\u003eair\u003c/sub\u003e(t+Δt) = αT\u003csub\u003eair\u003c/sub\u003e(t)+βSST(t)+γSL(t)), the model with only SST (red solid curve, T\u003csub\u003eair\u003c/sub\u003e(t+Δt) = αT\u003csub\u003eair\u003c/sub\u003e(t)+βSST(t)), sea level (green solid line, T\u003csub\u003eair\u003c/sub\u003e(t+Δt) = αT\u003csub\u003eair\u003c/sub\u003e(t)+βSL(t) ), or Nino3.4 index (yellow solid curve, T\u003csub\u003eair\u003c/sub\u003e(t+Δt) = αT\u003csub\u003eair\u003c/sub\u003e(t)+βNino3.4(t)), the persistence prediction (black dashed line, T\u003csub\u003eair\u003c/sub\u003e(t+Δt) = α+βT\u003csub\u003eair\u003c/sub\u003e(t)). The skills are cross-validated for the period of (a) January 1982 to December 2006 (b) January 2007 to December 2013 using analysis described in the text.\u003c/p\u003e","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-4577041/v1/c68921b74a9f45c1cd1579df.jpeg"},{"id":59749235,"identity":"226f041c-2ce7-49c5-91ca-4bc49fe3ec5b","added_by":"auto","created_at":"2024-07-05 19:17:07","extension":"jpeg","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":304196,"visible":true,"origin":"","legend":"\u003cp\u003eSame as Figure 5, but for Root Mean Square Error (RMSE) \u0026nbsp;as a function of forecast lead for predicted T\u003csub\u003eair\u003c/sub\u003e using model (2) (blue solid curve), only SST (red solid line), only sea level (green solid line), only Nino3.4 index (yellow solid line) and persistence skill (black dashed line) for the period of (a) January 1982 to December 2006 (b) January 2007 to December 2013.\u003c/p\u003e","description":"","filename":"floatimage6.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-4577041/v1/6b4a188f037c6ebc57f96980.jpeg"},{"id":59749237,"identity":"3ead4d97-f0b8-4e7a-8a30-511ed596daea","added_by":"auto","created_at":"2024-07-05 19:17:07","extension":"jpeg","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":465138,"visible":true,"origin":"","legend":"\u003cp\u003eSame as Figure 5, but for cross correlation for each target month during the (a) training (January 1940 to December 2007) and (b) validation (January 2008 to December 2023) period.\u0026nbsp; Here we used two predictors T\u003csub\u003eair\u003c/sub\u003e(t+Δt) = αT\u003csub\u003eair\u003c/sub\u003e(t)+βNino3.4(t)+γSL(t).\u003c/p\u003e","description":"","filename":"floatimage7.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-4577041/v1/67749a20d10ff86504a21c9d.jpeg"},{"id":61606641,"identity":"a7c636bb-e627-4238-aa1f-60829d835c91","added_by":"auto","created_at":"2024-08-01 22:01:43","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1651467,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4577041/v1/7f93d2ed-3914-4cde-9be1-0f9542f417a3.pdf"},{"id":59749234,"identity":"aa2a52d6-4aa4-4ef7-8ffb-0cc2c424b72a","added_by":"auto","created_at":"2024-07-05 19:17:07","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":80864,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterials20240613.docx","url":"https://assets-eu.researchsquare.com/files/rs-4577041/v1/06478eb430ddee593f7d5e06.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Using Ocean Data to Predict Monthly Air Temperatures of California's Coastal Cities","fulltext":[{"header":"Key Points","content":"\u003cp\u003e(1) A new predictor based on sea level and sea surface temperature anomaly is postulated for predicting air temperature at 4 main cities along California coast.\u003c/p\u003e\n\u003cp\u003e(2) Sea level and sea surface temperature can predict air temperature with 2-3 months in advance. The prediction capability is better during late summer/early fall.\u003c/p\u003e\n\u003cp\u003e(3) The predicting ability is limited within a 2-3 degree narrow coastal region and it decreases rapidly for inland stations.\u003c/p\u003e"},{"header":"Plain Language Summary","content":"\u003cp\u003eThis study postulates using Sea Level and Sea Surface Temperature Anomaly (SLA and SSTA) to predict the air temperature anomaly at 4 main coastal cities in California. Our results suggest that air temperature can be predicted by SLA/SSTA with 2-3 months in advance. The prediction capability is better during late summer/early fall, when the upwelling is strongest. The predicting ability is limited to a 2-3 degree coastal region, and it decreases rapidly further inland. Our work also suggests that it is very important to enhance the resolution of models near coastal region to improve seasonal prediction.\u003c/p\u003e"},{"header":"1. Introduction","content":"\u003cp\u003eAdvances in summertime temperature forecasts are important for planning in different economic sectors, especially the energy industry. This issue is especially important in California where the summer heavy loads from air conditioners, pumping water and other seasonal issues drive the peak energy demand 50% higher than that in winter (Alfaro, EJ et. al, 2004). Barnett and Preisendorfer (1987) explored the predictability of surface air temperature over the United States with advanced statistical techniques designed to maximize predictive skill and they checked the ability of a form of persistence, Sea Surface Temperature (SST) and Sea Level Pressure (SLP) fields. They also mentioned the potential role of El Ni\u0026ntilde;o-Southern Oscillation (ENSO), which was unclear at that time. Alfaro EJ et al. (2004) discussed a statistical prediction model based on canonical correlation analysis (CCA) and used the spring Pacific Sea Surface Temperature (PSST) as a predictor. They also mentioned that the Pacific Decadal Oscillation index (PDO) might be used as a potential predictor. But the correlation of their prediction model is not very high, the skill of seasonal average temperature (T\u003csub\u003emean\u003c/sub\u003e) and cooling degree days (CDD) prediction shows values greater than 0.4 (more than 20% of the variability predicted by model) in a large part of the region, while the prediction of warm extreme values provides acceptable values only for some locations in California (see Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e2\u003c/span\u003e in Alfaro EJ et al., 2004) and only about 30% of the stations, the correlation is greater than 0.4 (see Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e2\u003c/span\u003eb in Alfaro EJ et al., 2004) ; furthermore, their study was mainly focused on the warm temperature anomalies and their spatial distribution, without detailed information related to the main cities along the California coast.\u003c/p\u003e \u003cp\u003eDottori and Clarke (2009) show that low-frequency upper-ocean temperature anomalies near the California coast are mainly due to the advection of the mean temperature gradient by anomalous alongshore/vertical flow associated with coastal Kelvin waves and westward propagating large-scale Rossby waves. These waves are associated with coastal upwelling and sea level anomalies. Since both the anomalous vertical and alongshore currents are proportional to the time derivative of the interannual sea level (Dottori and Clarke, 2009), the upper ocean heat content anomaly (and SST) are closely linked to and in phase with the sea level anomaly. Thus, both of SST and sea level can be used to predict air temperature anomaly in the California coastal cities, as the predominant onshore winds along the coast can bring anomalous heat flux at the air-sea interface toward the continent (Morales-Acu\u0026ntilde;a et al., 2019).Sea level variations may represent the integrated effects of both remote and regional climate variability. Thus, a combination of SST and sea level may better capture the oceanic influences on land temperatures and serve as predictors. Land surface temperature predictions have been attempted by seasonal forecast models (Tan et al., 2014, Yang et al., 2022). We plan to assess the prediction skills with a simple statistical model in comparison with a fully coupled ocean-atmosphere-land model. Similar approaches have been adopted in other regions (Clarke and V. Gorder, 2001, 2003).\u003c/p\u003e \u003cp\u003eThe rest of the study is organized as follows. The data and methods used are summarized in Section \u003cspan refid=\"Sec2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. In Section \u003cspan refid=\"Sec5\" class=\"InternalRef\"\u003e3\u003c/span\u003e we show that sea level and sea surface temperature data can be used to predict air temperature 2\u0026ndash;3 months in advance, and it is better during late summer/early fall in 4 main cities, when the upwelling is strongest. However, the prediction is only limited within a 2\u0026ndash;3 degree coastal region; for stations further inland, the prediction skill declines rapidly. Moreover, a simple statistical model is set up, which can predict May to August T\u003csub\u003eair\u003c/sub\u003e better with 3.5 to 5 months in advance at San Diego Lindbergh Field. Section \u003cspan refid=\"Sec9\" class=\"InternalRef\"\u003e4\u003c/span\u003e discusses the potential link of SL (SST) prediction with ENSO dynamics and concludes this study with remarks.\u003c/p\u003e"},{"header":"2. Data and methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1. Dataset\u003c/h2\u003e \u003cp\u003eIn this study, we focus on four clusters of stations along the California coast (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e1\u003c/span\u003e, Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e). The \u0026ldquo;Revised Local Reference\u0026rdquo; in situ Sea Level (SL) data contains tide gauge observations, available from Permanent Service for Mean Sea Level website (PSMSL, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.psmsl.org/data/obtaining/\u003c/span\u003e\u003cspan address=\"https://www.psmsl.org/data/obtaining/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e ); the land based monthly air temperature (T\u003csub\u003eair\u003c/sub\u003e) station data is downloaded from Global Historical Climatology Network (GHCN)-Monthly Summaries ( \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.ncdc.noaa.gov/cdo-web/datatools/findstation\u003c/span\u003e\u003cspan address=\"http://www.ncdc.noaa.gov/cdo-web/datatools/findstation\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e ), which is a database for historical monthly temperature, precipitation, and snow records over global and areas. Monthly anomalous 1\u0026deg;x1\u0026deg; Sea Surface Temperature (SST) grid data (89.5\u003csup\u003eo\u003c/sup\u003eS to 89.5\u003csup\u003eo\u003c/sup\u003eN, 0.5\u003csup\u003eo\u003c/sup\u003eE to 359.5\u003csup\u003eo\u003c/sup\u003eE) from 1960 to current is available from the International Comprehensive Ocean-Atmosphere Data Set (ICOADS, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.esrl.noaa.gov/psd/data/gridded/data.coads.1deg.html\u003c/span\u003e\u003cspan address=\"http://www.esrl.noaa.gov/psd/data/gridded/data.coads.1deg.html\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). The 0.25\u0026deg;x0.25\u0026deg; topography data is from Joint Institute for the Study of the Atmosphere and Ocean (JISAO) data set (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://research.jisao.washington.edu/data_sets/elevation/\u003c/span\u003e\u003cspan address=\"http://research.jisao.washington.edu/data_sets/elevation/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e ) and Global Ocean Data Assimilation System (GODAS) data is from (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.cpc.ncep.noaa.gov/products/GODAS/\u003c/span\u003e\u003cspan address=\"https://www.cpc.ncep.noaa.gov/products/GODAS/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eFor each cluster, the in-situ SL data is available at the ocean station, air temperature data (T\u003csub\u003eair\u003c/sub\u003e) is available at each ocean station and four inland stations. The SST is based on ICOADS gridded data, and a grid closest to the in-situ ocean station is chosen. The correlation is estimated based on the overlap period of the data sets. For all the observed data, the monthly anomaly is calculated by removing the climatological mean of each calendar month throughout the record.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2. Cross-validated correlation skill\u003c/h2\u003e \u003cp\u003eThe data is separated into training and validation periods, here we used the period of January 1982 -December 2006 for training and 2007 January-December 2013 for validation. To evaluate the predictability of SLA and SSTA, here we used cross-validated correlation skill. As defined in Clarke and Van Gorder (2003), the correlations are estimated from the \u0026ldquo;Start Month\u0026rdquo; to a later month of Δt. For example, the correlation skill for a prediction of the March value of T\u003csub\u003eair\u003c/sub\u003e from January to a month of Δt\u0026thinsp;=\u0026thinsp;1.5months later, i.e., the middle of March.\u003c/p\u003e \u003c/div\u003e"},{"header":"3. Results","content":"\n\u003ch3\u003e3. 1. Prediction Results\u003c/h3\u003e\n\u003cp\u003eIn order to see the potential capability of using SST/SL to predict T\u003csub\u003eair\u003c/sub\u003e at the four clusters of land-based stations, the correlation between monthly anomalous SST vs T\u003csub\u003eair\u003c/sub\u003e and SL vs T\u003csub\u003eair\u003c/sub\u003e of the coastal stations of the four clusters are first assessed (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e2\u003c/span\u003e, also see the first column in Table S2). Our analysis suggests that the maximum correlation is 0.81 (R\u003csub\u003ecrit\u003c/sub\u003e(95%)\u0026thinsp;=\u0026thinsp;0.41) between SST and T\u003csub\u003eair\u003c/sub\u003e during September at San Diego Lindbergh Field (S\u003csub\u003e1,0\u003c/sub\u003e), and 0.86 (R\u003csub\u003ecrit\u003c/sub\u003e(95%)\u0026thinsp;=\u0026thinsp;0.39) between SL and T\u003csub\u003eair\u003c/sub\u003e during September. At Santa Monica Pier (S\u003csub\u003e2,0\u003c/sub\u003e), the maximum correlation between SST (SL) and T\u003csub\u003eair\u003c/sub\u003e also occurs during September, with a correlation of r\u0026thinsp;=\u0026thinsp;0.84 (R\u003csub\u003ecrit\u003c/sub\u003e(95%)\u0026thinsp;=\u0026thinsp;0.42) for SST and r\u0026thinsp;=\u0026thinsp;0.72 (R\u003csub\u003ecrit\u003c/sub\u003e(95%)\u0026thinsp;=\u0026thinsp;0.40) for SL. At Watsonville Waterworks (S\u003csub\u003e3,0\u003c/sub\u003e), the maximum correlation between SST(SL) and T\u003csub\u003eair\u003c/sub\u003e is slightly lower than S\u003csub\u003e1,0\u003c/sub\u003e and S\u003csub\u003e2,0\u003c/sub\u003e, which may be due to the topography of this region (see Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e1\u003c/span\u003e). At the last ocean station (S\u003csub\u003e4,0\u003c/sub\u003e, Eureka Weather Forecast Office Woodley Island), the maximum correlation between SST and T\u003csub\u003eair\u003c/sub\u003e occurs in August (r\u0026thinsp;=\u0026thinsp;0.87, R\u003csub\u003ecrit\u003c/sub\u003e(95%)\u0026thinsp;=\u0026thinsp;0.47), and happens in May between SL and T\u003csub\u003eair\u003c/sub\u003e (r\u0026thinsp;=\u0026thinsp;0.73, R\u003csub\u003ecrit\u003c/sub\u003e(95%)\u0026thinsp;=\u0026thinsp;0.45). Here the critical value is estimated based on Ebisuzaki et al. (1999) and the correlation between SST (SL) and T\u003csub\u003eair\u003c/sub\u003e is estimated based on the overlapped time period of two data sets. From this figure, we can see that during late summer/early fall (May to September), the correlation between SST, SL and T\u003csub\u003eair\u003c/sub\u003e are the highest. In other words, the influence of SST (SL) on T\u003csub\u003eair\u003c/sub\u003e appears to be strongest during late summer/early fall. Physically, this is mainly due to that the upwelling-favorable winds are the strongest during late summer/early fall, with the onshore component of the wind brings ocean induced temperature anomalies inland. At most coastal stations, the correlations between SL (or SST) and T\u003csub\u003eair\u003c/sub\u003e are the highest during this time of the year. Note here that, the correlation between SL (SST) and T\u003csub\u003eair\u003c/sub\u003e is largest for S\u003csub\u003e1,0\u003c/sub\u003e (San Diego Lindbergh Field) and decreases at the other three stations.\u003c/p\u003e \u003cp\u003eIn order to assess the potential lead months that SST (SL) can predict T\u003csub\u003eair\u003c/sub\u003e during late summer/early fall at each ocean station, lead correlations of SST with T\u003csub\u003eair\u003c/sub\u003e are calculated at each station (see Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e3\u003c/span\u003e). For example, Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e3\u003c/span\u003ea is the lead correlation between SST and T\u003csub\u003eair\u003c/sub\u003e during July-September. Our results suggest that SST can be used to predict T\u003csub\u003eair\u003c/sub\u003e during July-September for a lead of about 1 month (correlation decreases to about 0.4 after 1 month). However, the lead correlation between SL and T\u003csub\u003eair\u003c/sub\u003e suggests that SL can be used to predict T\u003csub\u003eair\u003c/sub\u003e by 2 months lead (correlation is about 0.4 after 2 months) in July-September. Physically, this is because the sea level is related to upper ocean heat content, reflecting both remote and regional climate drivers. Similarly, the lead correlations of SST (SL) with T\u003csub\u003eair\u003c/sub\u003e at the other three stations are estimated during the late summer/early fall (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e3\u003c/span\u003ec-h). The results suggest that at the three stations, SL can predict T\u003csub\u003eair\u003c/sub\u003e with a longer lead than SST. In other words, SL is a better predictor than SST at all four stations. More specifically, at S\u003csub\u003e2,0\u003c/sub\u003e, SST can predict T\u003csub\u003eair\u003c/sub\u003e by 1 month, however, SL can predict T\u003csub\u003eair\u003c/sub\u003e by 3months. At S\u003csub\u003e3,0\u003c/sub\u003e, SST can predict T\u003csub\u003eair\u003c/sub\u003e by 1 month, SL can predict T\u003csub\u003eair\u003c/sub\u003e by 3 months. At station S\u003csub\u003e4,0\u003c/sub\u003e, SST can predict T\u003csub\u003eair\u003c/sub\u003e by 1 month and SL can predict T\u003csub\u003eair\u003c/sub\u003e by 2 months. In summary, for all four cities, SST leads T\u003csub\u003eair\u003c/sub\u003e with a correlation of 0.5 or greater for some lead of about 1 month (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e3\u003c/span\u003e, (a), (c), (e)). In contrast, for SL, the lead increases to 2\u0026ndash;3 months (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e3\u003c/span\u003e, (b), (d), (f)).\u003c/p\u003e \u003cp\u003eTo explore the coastal region (coastal boundary layer) where the sea level can provide prediction of T\u003csub\u003eair\u003c/sub\u003e, near each alongshore ocean station, we picked up more stations inland as listed in Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e and calculated the maximum correlation for each calendar month. Figure\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e4\u003c/span\u003e shows that the correlation between monthly anomalous air temperature and SST (SL) falls with distance from the coast. Since the correlation falls very rapidly, such a marine boundary layer is difficult to be resolved by numerical models with a coarse resolution. Also note that the sudden drop in correlation for Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e4\u003c/span\u003ec and d (the dramatically lower correlation between SST and T\u003csub\u003eair\u003c/sub\u003e) might be associated with land topography, which blocks the influence of the ocean. The substantial correlation in Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e4\u003c/span\u003ed for Eureka far inland may have a topographic reason (see Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Table S2 further shows the maximum correlation at each station, which imply that the maximum correlations occur in late summer and fall (August-September) at all the stations (see second column in Table S2).\u003c/p\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e3.2. Cross-validated correlations\u003c/h2\u003e \u003cp\u003eBased on the precursor properties of T\u003csub\u003eair\u003c/sub\u003e, SL and SST, we write our predictor of T\u003csub\u003eair\u003c/sub\u003e(t\u0026thinsp;+\u0026thinsp;Δt) from month t as the linear sum of T\u003csub\u003eair\u003c/sub\u003e, SST (SL) as follows (also see Clarke and Van Gorder, 2003)\u003c/p\u003e \u003cp\u003eT\u003csub\u003eair\u003c/sub\u003e(t\u0026thinsp;+\u0026thinsp;Δt) = αT\u003csub\u003eair\u003c/sub\u003e(t)+βY(t) (1)\u003c/p\u003e \u003cp\u003ewhere T\u003csub\u003eair\u003c/sub\u003e is the raw air temperature value, Y(t) is the raw sea surface temperature value (or sea level value). In order to test the ability of (1) to predict T\u003csub\u003eair\u003c/sub\u003e(t\u0026thinsp;+\u0026thinsp;Δt) for various leads Δt, we determined the coefficients α and β by least-square fitting over a training (validation) period for each time Δt, here, we determined the coefficients over January 1982 to December 2006 (January 2007 to December 2013). Anomalies used during a given training period were based on seasonal cycles estimated during the whole period (January 1982 to December 2013). Cross-validated correlations for various lead times Δt at San Diego Lindbergh Field are shown in Table.S3. A lead time of \u0026frac12; month corresponds to (say) predicting the October value of T\u003csub\u003eair\u003c/sub\u003e given data to the end of September. Similarly, predicting the November T\u003csub\u003eair\u003c/sub\u003e value from the end of September would correspond to Δt\u0026thinsp;=\u0026thinsp;1.5months, etc. Figure.S1 shows that the model predicts T\u003csub\u003eair\u003c/sub\u003e in May to August with lead of 3.5 to 5 months.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e3.3. Improving the models\u003c/h2\u003e \u003cp\u003eIn order to improve the prediction, we combine sea level and sea surface temperature as shown in (2)\u003c/p\u003e \u003cp\u003eT\u003csub\u003eair\u003c/sub\u003e(t\u0026thinsp;+\u0026thinsp;Δt) = αT\u003csub\u003eair\u003c/sub\u003e(t)+βSST(t)+γSL(t) (2)\u003c/p\u003e \u003cp\u003eSimilar to (1), we determined coefficients α, β and γ by least-square fitting over a training period for each time Δt. Figure\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e5\u003c/span\u003e shows correlation skill at San Diego Lindbergh Field as a function of forecast lead for \u0026ldquo;prediction\u0026rdquo; of T\u003csub\u003eair\u003c/sub\u003e using model (2) (blue solid curve), the model with only sea level (green solid curve) or SST (red solid line). The skills are cross-validated for the period of January 1982 to December 2006 for training and January 2007 to December 2013 for testing, we can see that the sea level can be used as a predictor better than sea surface temperature, and combining sea level and SST is the best. The persistence skill (black dashed line) suggests that the T\u003csub\u003eair\u003c/sub\u003e can predict itself by about 1 month with a correlation of about 0.82. Note that the lead correlation based on calendar month (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e3\u003c/span\u003ea, b) suggests that comparing to SST, SL can predict T\u003csub\u003eair\u003c/sub\u003e by 2 to 3 months instead of 1 month. Here we chose San Diego Lindbergh Field, because Figs.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e2\u003c/span\u003e and \u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e3\u003c/span\u003e suggest that at this station, the link between SL and SST is the strongest among all four ocean stations. In order to test the predictability of ENSO, we also plotted the predicting result based on ENSO index (Nino3.4), the yellow line in Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e5\u003c/span\u003e, we can see that Nino3.4 has similar predictability to SST. Since the anomalous sea level at Santa Monica Pier is highly correlated at zero lag with the anomalous sea level at San Diego Lindbergh Field (not shown). We could also use the sea level at San Diego Lindbergh Field back to 1906 to predict the air temperature at the station near Santa Monica using Eq.\u0026nbsp;(1) with no SST term. We could even try to add in the SST term using the monthly SST results at Scripps Pier from Rasmussen et al. (2019). In order to evaluate the performance of air temperature prediction near station San Diego Lindbergh, we also estimated the root mean square error (RMSE) between observed T\u003csub\u003eair\u003c/sub\u003e and the predicted results based on SST, sea level, Nino3.4 index, and the combination of sea level and SST and the persistent skill (see Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e6\u003c/span\u003e). The results show that the RMSE for the SST-based prediction is larger than that based on SL, and the RMSE based on combining SL and SST is the smallest, which is consistent with Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e5\u003c/span\u003e. In Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e7\u003c/span\u003e, we show model skills stratified into different months, based on the model with Nino3.4 index and SL as predictors (T\u003csub\u003eair\u003c/sub\u003e(t\u0026thinsp;+\u0026thinsp;Δt) = αT\u003csub\u003eair\u003c/sub\u003e(t)+βNino3.4(t)+γSL(t) ), we can see that during May-July, the combination of Nino3.4 and SL can predict T\u003csub\u003eair\u003c/sub\u003e with a lead about 4 months.\u003c/p\u003e \u003c/div\u003e"},{"header":"4. Concluding remarks and discussion","content":"\u003cp\u003eUsing sea level, sea surface temperature and air temperature data available at ocean stations, a statistical model based on sea level and sea surface temperature was developed to predict the air temperature in the California coast region. The analysis shows that the sea level is a better predictor than SST and it also works better for late summer\\early fall than other seasons. Among four cities, the air temperature in San Diego Lindbergh can be predicted better. The prediction ability is limited to a 2\u0026ndash;3 degree narrow \u0026ldquo;boundary layer\u0026rdquo; along California coast, which is difficult to resolve for numerical model with coarse resolution. We also noticed that the sea level can predict itself very well; for example, at San Diego Lindbergh, sea level can predict itself by 2\u0026ndash;3 month lead with a correlation about 0.7 during July to October (see Figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e). The sea level and SST can be used as air temperature predictor is because physically, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\partial T{\\prime }/\\partial t\\)\u003c/span\u003e\u003c/span\u003e is proportional to \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\partial \\eta {\\prime }/\\partial t\\)\u003c/span\u003e\u003c/span\u003e and integration with respect to time. The relationship that \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\partial T{\\prime }/\\partial t\\)\u003c/span\u003e\u003c/span\u003e is proportional to \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\partial \\eta {\\prime }/\\partial t\\)\u003c/span\u003e\u003c/span\u003e follows from the balance at the sea surface (see Eq.\u0026nbsp;3.6 in Dottori and Clarke) and v' is proportional to \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\partial \\eta {\\prime }/\\partial t\\)\u003c/span\u003e\u003c/span\u003e from the freely propagating long wave Rossby wave balance. The San Diego sea level is basically in phase with the equatorial sea level, because even if the coastline were roughly 6000km from the equator to San Diego, it would only take about a month for the coastal Kelvin wave to cover this distance. Intra-seasonal Kelvin waves (periodicity about 30 days-75 days) do propagate along that coastline, and a monthly sea level would have some of that intra-seasonal signal contained in it. The 3-month running mean \"seasonal\" signal, which would filter out the intra-seasonal signal, may have a slightly different lag behavior. The lag due to a frictionless Kelvin wave propagation may be slight, anyway. Dissipation of the equatorial signal can also result in poleward propagation. Therefore, we expect that the persistence properties of \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\eta\\)\u003c/span\u003e\u003c/span\u003e at San Diego with the ENSO \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\eta\\)\u003c/span\u003e\u003c/span\u003e persistence properties at the equator.\u003c/p\u003e \u003cp\u003eAlso notice here, as discussed in the introduction, previous work has suggested the potential role of ENSO (Barnett and Preisendorfer, 1987) in predicting air temperature along California; however, the related dynamics are still unclear. For instance, ENSO generally has a peak in winter, however, our model suggests that SST and SL are highly correlated with T\u003csub\u003eair\u003c/sub\u003e during late summer/early fall. Basically, the ENSO dynamics can influence on air temperature prediction along the California coastal cities in the following two ways. Firstly, as shown in Dottori and Clarke (2009), the low-frequency upper-ocean temperature anomalies near the California coast are mainly due to the advection of the mean temperature gradient by anomalous alongshore (v\u0026rsquo;) and vertical (w\u0026rsquo;) flow associated with the westward propagating large-scale Rossby waves. These waves are associated with coastal upwelling and sea level anomalies, and they can be traced back to the coastal Kelvin waves remotely driven by the equatorial dynamics instead of local forcing (Enfield and Allen 1980; Chelton and Davis 1982; Kessler 1990). The upwelling is strongest during late summer/early fall, when the correlation between T\u003csub\u003eair\u003c/sub\u003e and SL is highest. Furthermore, the ocean dynamics basically sets the stage of T\u003csub\u003eair\u003c/sub\u003e. Along the coast of California, the climatological wind is equatorward, and it drives the cold coastal upwelling. The local wind stress can change due to variability in the atmospheric circulation. The other important factor is the ocean dynamics associated with ENSO, and these perturbations are mostly originated from the equatorial region and move poleward along the coast. The amplitude and duration of these Kelvin wave groups can enhance/reduce the coastal upwelling and change the local ocean condition in terms of SST and SL. The SST/SL anomaly for each month over the past 40 years based on GODAS data suggests that they are closely related to ENSO events, and they can be traced back to the equatorial region (not shown). Overall, our study helps to improve our understanding of the seasonal prediction along California coastal region and suggests that it is very important to enhance the resolution of models near coastal region to improve seasonal prediction.\u003c/p\u003e "},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe in situ Sea Level data is available from PSMSL (https://www.psmsl.org/data/obtaining/) . The land based monthly air temperature station data is downloaded from Global Historical\u003c/p\u003e\n\u003cp\u003eClimatology Network (GHCN)-Monthly Summaries (http://www.ncdc.noaa.gov/cdo-web/datatools/findstation). Monthly anomalous 1\u0026deg;x1\u0026deg;\u003c/p\u003e\n\u003cp\u003eSST data is available from the International Comprehensive Ocean-Atmosphere Data\u003c/p\u003e\n\u003cp\u003eSet (ICOADS) (http://www.esrl.noaa.gov/psd/data/gridded/data.coads.1deg.html).\u003c/p\u003e\n\u003cp\u003eThe topography data is from JISAO data set (http://research.jisao.washington.edu/data_sets/elevation/).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis research is supported by Dalian Maritime University Start Up funding and the Fundamental Research Funds for the Central Universities (3132024123).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026apos; contributions - provide individual author contribution\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eX.Z. and M.F. wrote the main manuscript text and X.Z. prepared figures 1-7. All authors reviewed the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgement:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe gratefully acknowledge support from Dalian Maritime University Start Up funding and the Fundamental Research Funds for the Central Universities (3132024123). The original idea was based on XZ\u0026rsquo;s PhD work at Florida State University. Prof. Allan J. Clarke first suggested to predict air temperature along California coast by using sea level data. Thanks also goes to Dr. Ruixin Huang, Mikhail Karpytchev, Takashi Mochizuki, whose comments have improved the original manuscript significantly.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAlfaro E J, Gershunov A, Cayan D R (2004) A method for prediction of California summer air surface temperature. \u003cem\u003eEOS Trans. AGU\u003c/em\u003e 85:553: 557\u0026ndash;558.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarnett T P, Preisendorfer R (1987) Origins and levels of monthly and seasonal forecast skill for united states surface air temperatures determined by canonical correlation analysis. \u003cem\u003eMon Wea Rev.\u003c/em\u003e 115: 1825\u0026ndash;1850.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChelton D, Davis R (1982) Monthly mean sea-level variability along the west coast of North America. J Phys Oceanogr. 12: 757\u0026ndash;784.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eClarke A J, Lebedev A (1999) Remotely Driven Decadal and Longer Changes in the Coastal Pacific Waters of the Americas. Journal of Physical Oceanography. 29: 828\u0026ndash;835.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eClarke A J, Dottori M (2008) Planetary Wave Propagation off California and Its Effect on Zooplankton. \u003cem\u003eJ\u003c/em\u003eournal of Physical Oceanography. 38: 702\u0026ndash;714.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eClarke A J, Van Gorder S (2001) ENSO prediction using an ENSO trigger and a proxy for western equatorial Pacific warm pool movement. Geophysical Research Letters. 28(4): 579\u0026ndash;582. doi:10.1029/2000GL012201.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eClarke A J, Van Gorder S (2003) Improving El Ni\u0026ntilde;o prediction using a space-time integration of Indo-Pacific winds and equatorial Pacific upper ocean heat Content. Geophysical Research Letters. 30(7), doi:10.1029/2002GL016673.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDottori M, Clarke A J (2009) Rossby Waves and the Interannual and Interdecadal Variability of Temperature and Salinity off California. Journal of Physical Oceanography. 39: 2543\u0026ndash;2561.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEbisuzaki W (1997) A method to estimate the statistical significance of a correlation when the data are serially correlated. J Climate. 10: 2147\u0026ndash;2153.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEnfield D B, Allen J S (1980) On the structure and dynamics of monthly mean sea level anomalies along the Pacific coast of North and South America. Journal of Physical Oceanography. 10(4): 557\u0026ndash; 578, doi:10.1175/1520-0485(1980)010\u0026lt;0557:OTSADO\u0026gt;2.0.CO;2.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMorales-Acu\u0026ntilde;a E., Torres C R, Linero-Cueto J R (2019) Surface wind characteristics over Baja California peninsula during summer. Reg. Stud. Mar. Sci., 29, p. 100654, 10.1016/j.rsma.2019.100654.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKessler W S (1990) Observations of long Rossby waves in the northern tropical Pacific. Journal of Geophysical Research: Oceans. 95 (C4): 5183\u0026ndash;5217.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePhan-Van T, Nguyen-Quang T, Trinh-Tuan L ,et al. Seasonal Prediction of Summer near-surface Air Temperature for Vietnam using RegCM4.2[C]//The Third International MAHASRI/HyARC Workshop on Asian Monsoon and Water Cycle.2013.DOI:10.5930/issn.1994-4683.2014.07.113.p168-171.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRasmussen L L, Carter M L, Flick R E, Hilbern M, Fumo J T, Cornuelle B D, et al. (2020) A century of Southern California coastal ocean temperature measurements. Journal of Geophysical Research: Oceans. 125, e2019JC015673. https://doi.org/10.1029/2019JC015673.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYang Y, Sun W, Zou M, Qiao S and Li Q (2022) Multi-model seasonal prediction of global surface temperature based on partial regression correction method. Front. Environ. Sci. 10:1036006. doi: 10.3389/fenvs.2022.1036006.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Ocean Dynamics, Air Temperature Prediction, California's Coastal Cities, Sea Level Anomaly","lastPublishedDoi":"10.21203/rs.3.rs-4577041/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4577041/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eAir temperature anomalies in the narrow California coastal boundary layer is closely associated with the adjacent coastal sea level variability. Coastal sea levels, proxies of upper ocean heat content, are regulated by alongshore winds and poleward propagating coastal Kelvin waves. Increased upper ocean heat content or higher sea level can enhance oceanic heat release to the atmosphere which will be carried inland due to the prevailing onshore winds. Based on these physical processes, sea level anomaly (SLA) and sea surface temperature (SSTA) are combined and used as a predictor here to forecast air temperature anomaly at 4 main cities along the California coast. Results show that air temperature in these land-based stations can be predicted by SLA and SSTA with 2\u0026ndash;3 months in advance. The prediction capability is better during late summer/early fall when the upwelling is strongest. The predicting ability is limited within a 2\u0026ndash;3 degree narrow coastal region, and it decreases rapidly for inland stations. Our work also suggests that it is very important to enhance the resolution of models near coastal region to improve seasonal prediction.\u003c/p\u003e","manuscriptTitle":"Using Ocean Data to Predict Monthly Air Temperatures of California's Coastal Cities","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-07-05 19:17:02","doi":"10.21203/rs.3.rs-4577041/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"46a09279-142c-4a61-9d03-bd69f9ca03e6","owner":[],"postedDate":"July 5th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2024-08-01T21:53:35+00:00","versionOfRecord":[],"versionCreatedAt":"2024-07-05 19:17:02","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-4577041","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4577041","identity":"rs-4577041","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}